TCS Daily

Veritas No Longer

By John Luik - June 21, 2005 12:00 AM

"These results are completely misleading"
-- Walter Willett in "Obesity: Time bomb or dud?", USA Today May 26, 2005

Ever since the Centers for Disease Control and Prevention's beliefs about the connection between being overweight, obese and dying were turned on their head earlier this spring, the true believers in the fat=death faith have been plotting how to get those embarrassing news conferences -- with CDC Director Julie Gerberding trying to explain how the CDC got its figures and its claims so badly wrong -- off the evening news. The new strategy for bringing the media back on their side -- an indispensable requirement for scaring Americans about being fat -- appears to be to move the focus of the campaign about the dangers of being overweight away from the discredited CDC to the purer intellectual pedigree of Harvard. After all, we may not trust what the government tells us about the dangers of being fat, but who can doubt Harvard?

The Harvard people, as one might expect, have been quick off the mark with a "Symposium on Overweight, Obesity and Mortality" organized by Walter Willett of Harvard's Department of Nutrition and one of the high priests of the fat=death faith. Willett, for instance, has argued that there is a "strong international consensus among scientists that overweight and obesity are major contributors to morbidity and mortality" and he attacked the CDC's lower obesity mortality estimates as "completely misleading".

The two hour symposium included faculty members from Harvard and the University of Texas Southwestern Medical Center, the Vice President of Epidemiology and Surveillance Research of the American Cancer Society and Donna Stroup from the CDC. Dr. Stroup -- one of the authors of the CDC's now discredited 2004 study claiming 400,000 annual obesity deaths -- seemed a strange choice given that the CDC now puts obesity mortality at 26,000 and an internal review of the paper written by Stroup and her CDC colleagues was highly critical of her work. Perhaps Stroup, whose role at the symposium was to "provide perspective on bodyweight from CDC studies" -- an odd assignment for someone who got the body count from obesity wrong by about 325,000 -- was there to provide a bit of balance to the program. She certainly didn't do anything to defend her CDC colleagues as at one point she said of her Harvard friends, who were busy trashing her own employer's study, "I would not disagree with any of my colleagues on the program."

In any event, balance was hardly what the Harvard symposium was about. (Given recent events at Harvard, this might not seem surprising.) Perhaps a more apt title might have been the less academically polite but more honest: "Why The CDC Doesn't Know What it is Talking About: How Your Fat Will Kill You", for beginning with Professor Willett, the task of each of the speakers was to convey, in a surprisingly direct way, why the CDC was so badly wrong. According to the Harvard folk, the CDC lower obesity mortality numbers were wrong for two major reasons: the CDC data was contradicted by other, better data and the CDC incorrectly analyzed and interpreted its own data.

Better Data

Leading the attack on the basis of better data was Harvard's JoAnn Manson, who authored a 1995 study "Body Weight and Mortality Among Women" based on the Nurses Health Study, a cohort of 121,700 female registered nurses from 30-55 years old, whose lives and deaths were followed from 1976-1992. Manson's study is perhaps the most frequently cited work to support the claim that even small amounts of extra weight are a significant health risk. Commenting on the Manson study's link between obesity and mortality, symposium participant Frank Hu from Harvard noted that as the BMI increased, "the death rate increases dramatically."

But that isn't what Manson's data actually say. Indeed, one of the disturbing things about the obesity studies that are most frequently used as proof texts in support of the claim that extra pounds translates into an early death is how different the data turns out to be from the claims made about that data by the study's authors or by the screaming headlines of the fat-obsessed media. What Manson's data show -- as opposed to her, Hu's or Harvard's spin -- is that there is no statistically significant increase in mortality risk for all women with BMI's ranging from 19.0-31.9, a BMI range that includes normal, overweight and obese. In other words the claim that just a small amount of extra weight increases risk of death is simply not true since even women with BMI's of 29-31.9 did not have a statistically significant increase in mortality risk.

Furthermore, much as the recent CDC study discovered, the lowest mortality risk was found for women who were mildly overweight with BMI's of 25-26.9. Rather than providing the basis for attacking the CDC, Manson's study actually corroborates its findings. Again, even women with the highest weights, BMI's greater than 32 had a RR of only 1.5 which is an extraordinarily weak association that brings into doubt the connection between obesity and death.

Responding to the Harvard claims about the Nurses study, Dr. Katherine Flegal, the author of the revised CDC study which has so upset Harvard, points out that the Manson analysis of the Nurses study excludes large numbers of deaths, 90% in one analysis, to produce its figures. These customized exclusions, often referred to as "data trimming", can produce about any result that the researcher wants and are unacceptable in clinical trials. What's more, the study did not use actual measured weights, but relied on the participants' self-reported estimates, unlike the CDC data. Most importantly, the study looks only at nurses, and not only nurses, but white, middle class nurses, hardly a representative sample of Americans. The CDC data, based on the National Health and Nutrition Examination Surveys (NHANES), is a much more representative sample of the national population.

Also joining in the attack on the new CDC obesity numbers was Michael Thun from the American Cancer Society. Don't believe those new CDC numbers, opined Dr. Thun, "Overweight people do have higher all-cause death rates" and this "increase in mortality includes major cancers..." But that's an odd claim coming from Dr. Thun whose own 2003 study on cancer and mortality tells a different story. In overweight men, being overweight was protective against cancer -- again a similar finding to the CDC's about the protective virtues of being overweight. In all populations, according to the Thun study, relative risks were less than 2, suggesting a lack of association between cancer and obesity. And this data shouldn't surprise Dr. Thun inasmuch as the Seven Countries Study involving 13,000 men over the last 40 years found that "... the risk of dying from cancer decreased with increasing relative weight."

Better Interpreted Data

The Harvard true believers had a second line of attack on the CDC's sharply lower obesity deaths, however. It wasn't, they contended, simply that the CDC's numbers were contradicted by other, better studies -- just coincidentally done by Harvard researchers -- but they were contradicted by their own data. In other words, the Harvard people took the CDC's data, reanalyzed that data and found quite amazingly that the CDC had gotten their analysis wrong once again. The problem, according to Dr. Hu, is that the CDC people included smokers and people who might have been ill during the study, thus injecting bias into the results by artificially increasing mortality for those with lower weights -- since both the ill and smokers weigh less.

Responding to these claims, the CDC's Dr. Flegal -- who was not at the Harvard symposium devoted to attacking her work -- noted that she didn't "have a problem with people at a conference talking about their data, but I do have a problem with their talking about our data and saying we should have found the same things that they found." Flegal notes that the Harvard criticisms of her numbers are simply not true since she and her colleagues at the CDC looked at the data in five different ways: thin individuals with stable weight for three years, five years, and ten years, thin smokers and thin nonsmokers. But in each of these samples the risks of the very thin and the very obese were the same. Moreover, Flegal said that she looked at the data both including and excluding current and former smokers and also ill people with chronic diseases. In each instance, however, the results were consistent: there was no increased mortality risk associated from being overweight and being obese up to a BMI of 35.

What Flegal was too gracious to point out was that her paper was actually published in a peer- reviewed journal unlike the off the cuff and inaccurate criticisms from Harvard that were delivered by sound bite at a news conference. Nor did she point out that the CDC's new obesity numbers are supported by dozens of other international studies involving hundreds of thousands of subjects done over the last forty years.

Of course the real purpose of the Harvard "symposium" had little to do with serious science as opposed to attempting to get the junk science obesity crusade back on track. As the CDC's Donna Stroup put it in perhaps too candid a fashion:

        "We are all trying to get to the same side of this issue -- that obesity is 
        a very important problem.... My point is that the differences are less 
        important than getting on with interventions that we know work."

But with respect, Dr. Stroup, we are not all on the same side of this issue for the entire point of your colleagues' study is that the evidence shows that obesity is NOT a "very important problem" and that means that we should NOT be "getting on with interventions" of any kind. At the end of the day about the only thing misleading was not, as Walter Willett claimed, the new CDC numbers, but rather the Harvard take on the "obesity crisis".

The author is a writer living in Canada. He is writing a book about health policy.


TCS Daily Archives